Advancing Pediatric Fracture Detection: A Novel YOLOv9 Architecture for Enhanced X-ray Image Analysis

Traditional YOLO models often struggle with the automated detection of pediatric wrist fractures in X-ray images, particularly in achieving high classification accuracy with complex and subtle fracture patterns due to their limited feature extraction capabilities. In response to these challenges, this research significantly enhances the YOLOv9 framework by integrating a specialized branch network that focuses on refining feature extraction and fracture classification. This innovation leads to substantial gains in detection performance, achieving a mean average precision (mAP) of 71.8% and an mAP50-95 of 43.4%. Our results demonstrate the model’s improved ability to identify intricate and subtle fracture patterns, offering a promising solution to reduce radiologist workload and accelerate diagnostic processes in clinical settings.